12/23/2008 09:52 612.455.3801 HSML (JLS) Page 3/16
Application Number 10/758815
Response to the Office Action dated September 24, 2008
REMARKS
Applicants request reconsideration of the rejection in view of these remarks.
Claims 1-1 5 and 18-29 are pending. Claims 24-29 are withdrawn from consideration.
The rejection under 35 K&C U 03(a)
Applicants maintain their traversal of the rejection of claims 1-3, 5-11, 13, and
16-23 as being obvious over Kidoguchi '586, the rejection of claim 4 as being
obvious over Kidoguchi '586 in view of Koike '770, the rejection of claims 12 and 14
as being obvious over Kidoguchi '586 in view of Sarayama '663, and the rejection of
claim 15 as being obvious over Kidoguchi '586 in view of Sarayama '663 and
D'Evelyn '434. Particularly, Applicants traverse the rejection's assertion that the
formation of a Group III nitride layer having a cycle of gaps of at least 100 \xm is
merely discovering an optimum value of a result effective variable that involves only
routine skill in the art.
Applicants have explained in the previous amendment and response mailed
April 11, 2008 that forming a seed crystal substrate with a cycle of gaps of at least
100 urn was not thought to be possible at the time of the invention, let alone involving
"only routine skill in the art to optimize a value."
Applicants further submit two research papers to provide evidence and
scientific reasoning that forming Group III nitride crystals on a Group III nitride layer
having a cycle of gaps of at least 100 urn could not have been considered a problem
of mere optimization and was not within the purview of one of ordinary skill in the art
at the time of the present invention. One paper, Ishibashi et al., Jpn. J. AppL Phys.
Vol.42 (2003) pp. L 1248-1251 explains the problems of tilt and distortion when
growing GaN crystals on a seed substrate having gaps using a vapor growth method.
This is shown in, for example, Figures 2(a) to 2(c) of Ishibashi and especially the
explanatory diagrams below the figures. Kidoguchi '586 as stated in column 17, lines
34-37 discloses a cycle of gaps of 1 8 \im. Ishibashi teaches that the problems of tilt
and distortion at this cycle of gaps of 18 (am, the same periodicity of Kidoguchi '586,
2
PACE 3/16* RCVD AT 12/23/2008 10:52:25 AM [Eastern Standard Time] * SVR:USPTO-EFXRF-5/8 » DNIS^maOO • CSID:61 2-455-3801 * DURATION (mm-ss):05-40
12/23/2008 09:52
612.455.3801
HSML(JLS)
Page 4/16
Application Number 10/758815
Response to the Office Action dated September 24, 2008
are very evident. Ishibashi, moreover, discloses that as the width of the wing over the
spacing of gaps increases, i.e., as the cycle of gaps become larger, the tilt angle
increases. As the tilt angle increases, the wings grown over the respective convex
portions, i.e., the gaps, become resistant to coalescing with each other. Ishibashi
establishes that growing crystals on a seed substrate with a gap spacing of 1 8 \xm is
extremely difficult and makes no mention of growing crystals at a larger gap cycle.
Thus, at the time of the invention, not only do Kidoguchi '586 and Ishibashi not
provide any suggestion or reason to try to grow crystals on a Group III nitride layer
having a cycle of gaps of at least 100 fim, these references, especially Ishibashi, teach
away from the claimed invention.
The second paper, Nataf et al., Journal of Crystal Growth 192 (1998) pp. 73-
78 also describes growing GaN layers on subtrates using vapor-phase epitaxy. When
GaN crystals were grown on a seed substrate with masks (gaps) of 5 nm and windows
(exposed portions of the GaN crystals) of 5 urn by a vapor growth method, plate-
shaped GaN crystals were obtained, provided the spacing region was not more
than 10 ^im, see page 75, column 1, and Figure 3a (emphasis added). Figure 3b
explains that for a dielectric width larger than 10 ^m, hexagonal holes appeared even
for growth times as long as three hours, evidencing the difficulty for GaN bridges to
completely link perpendicular to the stripes direction, see page 75, column 1, and
Figure 3b. Moreover, when a seed substrate with masks (gaps) of 1 00 nm and
windows (exposed portions of the GaN crystals) of 10 jim were used, GaN grown
from respective windows did not coalesce with each other at all; Figure 3c illustrates
this phenomenon. Thus, Nataf et ah describe that there was not an expansion of a
lateral plane (wing) leading to coalescence and plate-shaped GaN crystals were not
obtained, see p. 75, left column, lines 1 to 6, and Figures 3b and 3c. In other words,
Nataf et al. teach that growing Group III nitride crystals on a Group III nitride layer
having a cycle of gaps of at least 100 ^im could not be done.
These two references clearly demonstrate that, at the time of the invention,
growing crystals on a seed substrate with a spacing of gaps of at least 100 iim was
3
PACE 4/16* RCVD AT 12/23/2008 10:52:25 AM [Eastern Standard Time] ■ SVR:UBPTO-EFXRF-5/B * DN1B:2738300* CSID:81 2-455-3801 ■ DURATION (mm-ss):05-*0
12/23/2008 09:52 612.455.3B01
HSML (JLS)
Page 5/16
Application Number 10/758815
Response to the Office Action dated September 24, 2008
extremely difficult, if not impossible, in light of the technical common knowledge at
the time. A prima facie case of obviousness can be rebutted by showing that the art, in
any material respect, teaches away from the claimed invention. Applicants assert that
a teaching of a cycle of gaps of 5-20 jam, as in Kidoguchi 6 586, Ishibashi, and Nataf,
teach away from a cycle of gaps of at least 100 \xm, as required by claims 1,12, and
18.
Further, none of Koike fc 770, Sarayama '663, and D'Evelyn '434, discloses,
describes, suggests or provides a motivation for growing group III nitride crystals on
a seed substrate with a cycle of gaps of at least 100 ^m. Here, the rejection has not
pointed to any teaching in the cited references, or provided any explanation based on
scientific reasoning, that would support the conclusion that those skilled in the art
would have considered it obvious to "optimize" the cycle of gaps to at least 100 jam,
as required by claims 1,12 and 18. On the contrary, Applicants have presented
evidence that teaches growing group III nitride crystals on a seed substrate with a
cycle of gaps of at least 100 \im could not be done. Ex parte Whalen II , BPAI (July
23, 2008) at 14.
Applicants assert that independent claims 1,12 and 1 8 are not obvious and are
allowable. Claims 2-11, 13-15 and 19-23 are also allowable at least by virtue of their
dependence upon claims 1 and 18. Applicants do not concede the correctness of the
rejection.
4
PACE 5/16 ■ RCVD AT 12/23/2008 10:52:25 AM [Eastern Standard TlmeJ * SVR:UBPTO-EFXRF-5/8 * DNIS:2738300 • CSID:612-455-3801 * DURATION (mm-ss): 05-40
12/23/2008 09:52 612.455.3801
HSML (JLS)
Page 6/16
Application Number 10/758815
Response to the Office Action dated September 24, 2008
Should there by any remaining issues that could be easily resolved by
telephone, the Examiner is invited to telephone Douglas P. Mueller at 612.455.3804.
53148
PATENT TRADEMARK OFFICE
Dated: December 2-3 . 2008
Respectfully submitted,
HAMRE, SCHUMANN, MUELLER &
LARSON, P.C.
P.O. Box 2902
Minneapolis, MN 55402-0902
(<n2W455-3800
Douglas P. Mueller
Reg. No. 30,300
DPM/KO/ad
PACE 6(16 • RCVD AT 12/23/2008 10:52:25 AM (Eastern Standard Tlmel * SVR:USPTO-EFXRF-5/8 * DNIS:273830O * CSID:612-*55-3801 * DURATION (mm^s):05-4O
12/23/2008 09:52
612.455.3801
HSML (JLS)
Page 7/16
Jpn. J. Appl. Phy*. Vo!. 42 (2003) pp. L 124&-4. 1251
Pari 2. No. 10H, 15 October 2003
(£2003 The Japan Society of Applied Physics
Study on Deformations and Stress Distributions
in Air-Bridged Lateral-Epitaxial-Grown GaN Films
Akihiko Ishibashi*, Gaku Sucjahara, Yasutosbi Kawaguchi and Toshiya YOKOGAWA
Advanc ed Technology Research LabomtorUs, Matsushita Elecflic Industrial Co., Ltd., S t- 1 Yagumo-Nakamachi. Motiguchh Osaka 570-850 J, Japan
(Received July 29, 2003; accepted August 25, 2003: published October 8, 2003)
Two-dimensional deformations and stress distributions in air-bridged lateral -epitaxial -grown GaN (ABLEG-GaN) films have
been studied by atomic force microscopy (AFM), two-dimensional finite element method (FEM) analysis, and micro-Raman
spectroscopy. The ABLEG-GaN wings slightly till, and the direction of the wing tilt changes when the wings coalesce with
each other. After coalescence of the wings, the tilt angle decreases with increasing film thickness. By FEM analysis and Raman
spectroscopy, it has been revealed that the deformation of the wings originates from the distributions of thermal stress due to
large mismatch of the thermal expansion in the GaN seed layer and in the sapphire substrate. The wing deformation is
suppressed with increasing film thickness, since the stress distribution becomes more uniform.
[DOI: 10.I143/JJAP.42.LI248]
KEYWORDS: GaN, air bridge, thermal stress, micro- Raman, finite element method
Group III nitrides are highly promising for applications in
blue and ultraviolet optoelectronic devices and high- temper-
ature, high-power transistors. Recent studies on the various
selective area growth (SAG) techniques, such as epitaxial-
lateral-overgrowth (ELO) and Pendeo-epitaxy of group HI
nitrides on sapphire and SiC substrates via metal -organic
vapor phase epitaxy (MOVPE) have shown that they were
extremely effective in reducing the threading dislocation
(TD) density and markedly improved the performance of the
devices, such as violet laser diodes. , ~ 5> However, there is a
problem in that the ELO-GaN regions, hereafter called
"wings**, till cry stall ographically (about 1°). It is supposed
that the crystallographic tilt of the wings originates from the
stress at the interface between the wings and the mask, and
the thermal stress due to the mismatch of thermal expansion
between the GaN layer and a substrate, 6) The wing lilt
induces the dislocations near the coalescence region of the
7)
wings.
We have developed an advanced ELO structure, namely,
"air-bridged lateral epitaxial grown (ABLEG)"-GaN, 89>
which has no contact between the wing and the mask, as
shown in Fig. 1 . Since the wings in the ABLEG-GaN have no
contact with the masks, there is no stress in the interface
between the wings and the masks. Therefore, the wing tilt
markedly decreases in the ABLEG-GaN and the TD densities
are reduced not only in the wing region but near the
coalescence boundary beiween the wings. The tilt angle is
approximately 0.T, and the TD density is l^cm" 2 in the
ABLEG-GaN wings. It is supposed that the wing tilt is due to
the thermal stress generated between the GaN-seed layer and
the sapphire substrate, not the stress in the interface between
the wings and masks. The local stress analysis of the ELO-
GaN deposited on SiOi masks was reported. 10 * Though the
strain and crystallographic tilt in uncoalesced GaN layers
grown by mask less Pendeo-epitaxy have been studied, m
there are few reports on the quantitative analysis of the
deformation induced from the thermal stress in the coalesced
ELO-GaN.
In this work, we have studied the wing deformations
resulting from the thermal stress distributions in the ABLEG-
GaN tilnus on the sapphire substrates by atomic force
air-bridged structure
r A
seed wing
coalescence
boundary
air gap Void
"E-mail address: isibasi.ukihiko@jp.panasonic.com
Fig. t. Schematic cmsa-scctionul diagram of the ABLEG-GaN.
microscopy (AFM), ,the finite element method (FEM) based
on the two-dimensional orthotropic linear elastic theory, and
micro-Raman spectroscopy. Tt was found that the calculation
results on the deformation of the wings in the ABLEG-GaN
are in good agreement with the experimental results.
Figure 1 shows the schematic cross-sectional diagram of
the ABLEG-GaN. A 1 .2-|im-thick (0001) GaN layer is grown
on the (0001) sapphire substrate with a 20-nm-thick low-
temperature (LT) GaN buffer layer by low-pressure MOVPE.
The GaN film is grooved to form ridge snipes along the
< 1 1 0l)>cmN direction. The width und the period of the ridge
stripes are 3 and 15 Mm, respectively. Next, the 10-nm-thick
silicon nitride mask is deposited on the grooved GaN film.
After that, only the silicon niuride mask on the top of the ridge
stripes is removed, and the silicon nitride mask on the
sidewalls and bottoms of the grooved GaN film is left. The
ABLEG is performed on the exposed (0001) GaN top surface
which is a seed crystal for regrowth. When the coalescence of
the wings occurs, the air-bridged structure is constructed. A
smooth surface is obtained not only in the wing region, but in
the vicinity of the coalescence boundary. The tapered voids
arc formed below the coalescence boundary surfaces. Some
details of the growth and process conditions are reported in
refs. K and 9.
The surface conditions and deformations of the wings are
characterized by AFM. The AFM is carried out in the tapping
mode using a Digital Instruments Dimension 3100 scanning
probe microscope. The crystallographic lilt is characterized
by X-ray diffraction {XRD) measurements.
PACE 7/16 * RCVD AT 12/23/2008 10:52:25 AM [Eastern Standard Time] * SVR:USPTO-EFXRF-5/B * DNIS:2738300 * CSID: 612-455-3801 * DURATION (mm-SS):05-40
12/23/2008 09:52
612.455.3801
HSML (JLS)
Page 8/16
The wing deformations resulting from the thermal stress
distributions in the ABLEG-GaN films on the sapphire
substrates have been simulated by FEM based on the two-
dimensional orthotropic linear elastic theory. 12 * We assumed
that the thermal stress is only due to the mismatch of thermal
expansion coefficient {«) between GaN. and the sapphire
substrate during the cooling process after epitaxial growth.
The temperature is changed from 1000°C to 25°C during the
cooling process. The components of the elasticity theory are
the mismatch of a between GaN and the substrate and a set of
moduli of elasticity. We used the moduli of elasticity for GaN
and the sapphire in refs. 13-15.
The stress distributions are characterized by a frequency
shift in the micro-Raman spectroscopy measurements. In
these measurements, light from the 514.5 nm line of an Ar + -
ion laser is focused on the (1100)^ surface of the cleaved
ABLEG-GaN at room ternperature. The diameter of the
focused area is below I pm. Tn this backscattering geometry,
the A. i (TO), Ej (TO), and Ez modes are allowed on the basis
of the selection rules.
Figure 2 shows the cross-sectional scanning electron
microscopy (SEM) image and the AFM profiles of the
surface in the ABLEG-GaN, whose wings are uncoalesced
(aj, and coalesced (b)-{c), respectively. In the case of the
uncoalesced wing, the height of the wing edge is lower than
that of the center of the wing. The till angle f?i, which is
defined as arctan (fyM) in Fig. 2 (a), is about -0.06 y
0.02°, where the minus sign denotes the <000I>gbN
direction, [n case of the coalesced wings, the coalescence
boundary is higher than that in the seed region. The tilt angle
f% is 0.17'' -v 0.2 P for the film thickness of 2.5 um, and 0* is
0.02 ' ~ 0.06 c for the film thickness of 8.0 um. Therefore, the
till angle decreases with increasing film thickness.
In order to determine the origin of the wing deformation.
wing I wing
! i J tl»=0.0B3 o
tllt=Q.035°
•1000-500 0 500 1000
<D(arc«ec)
2.5-um-thlck GaN
(a)
-1000-500 0 50O 1000
a> (arcsec)
B,PrWlVttllCK QflN
(b)
Fi£. 3. X-ray rocking curves <XRCs) of the <00U2) u> scan for v? = 0°.
where <p is the angle between the rotation axis of the a> scan and (he stripe
direction.
XRD measurements are performed, where the direction of the
incident X-ray beam is perpendicular to the stripe. Figure 3
shows X-ray rocking curves (XRGs) of the (0002) to scan for
(p = 0°, where the azimuth <p is the angle between the rotation
axis in the a> scan and the stripe direction. Three diffraction
peaks are observed, that is two satellite peaks and one small
peak at the center between the satellite peaks. The two
satellite peaks are located ±0.083° (±0.035°) from the center
peak for the 2.5-jim (8,0-u.m)-thick GaN film. The till angle
decreases with increasing film thickness. For <p = 90 \ only
the single peak is observed for any thickness. Therefore, the
wing deformation originates from the crystallographic tilt of
the c-axis perpendicular to the stripe.
Since the thermal expansion coefficient of GaN is smaller
than that of sapphire, the compressive stress is prodaced in
PACE 8/16 * RCVD AT 12/23/2008 10:52:25 AM [Eastern Standard Time) • SVR:USPTO-EFXRF-5/8 " DNIS:2738300* CSID:612-455-3801 * DURATION (mm^s):05-40
12/23/2008 09:52
612.455.3801
HSML (JLS)
Page 9/16
L 1250 Jpn. J. Appl. Phys. Vol. 42 (2003^ Pi. 2, No. 10B
A. IsHlBASHf cr a!.
(a)
(b)
(c)
Fig. 4. Calculated thermal stres* distributions along the jr-a*is (a, j before coalescence of the ™gs (aX and after coalescence of the
wings (b>(c). respectively.
the epitaxial GaN layer on the sapphire substrate when the
sample is cooled from the growth temperature to the room
temperature. The stress distribution produces elastic defor-
mation of the GaN layer. In ABLEG-GaN, since the wings
are located above the air gap, the elastic deformation can
easily occur. We have simulated the elastic deformation by
FEM.
The calculated thermal stress distributions along the jr-axis
(tr„) are shown, for the 3-um-tbick wing (uncoalesced) in
Fig, 4 (a), the 3-u.m-thick wing (coalesced) in Fig. 4 (b), and
the 6-uin-thick wing (coalesced) in Pig. 4 (c). The shape is
modified, on the condition that the nodal displacement along
the y direction (DY) is multiplied by 16, in order to make it
easy to understand the deformation. Before coalescence, the
compressive stress (about 300 MPa) exists in ihe GaN seed
layer due to the large mismatch of the thermal expansion
coefficient. In the seed region of ABLEG-GaN, a large
compressive stress exists at the bottom, while a small tensile
stress exists near the surface. In the wing region, the stress is
nearly free, As a result, the wing bends slightly downwards.
Just after coalescence, the compressive stress (about
300 MPa) exists in the GaN seed layer, similar to that before
coalescence. In the wing region, a large compressive stress
concentrates at the coalescence boundary region and the
tensile stress exists at the bottom surface of the wings. As a
result, the wings bend and the coalescence boundary becomes
higher than the seed region. The till angle of the wings, for
the thickness of 3 urn, is about 0.2°, which is in good
agreement with the experimental result. For the thickness of
6u\m, a larger compressive stress concentrates at the
coalescence boundary region compared with other regions.
However, the stress becomes more uniformly distributed
along the x-direction, compared with that lor the thickness of
Sunl-
it is supposed that the wing deformations originate from
the stress distributions and the relaxation in the air gap, the
void, and the surface. There is highly compressive stress in
the GaN seed layer, which brings about compressive stress in
the wings. Since the compressive stress can relax in the x-
direction on both sides of the void, the wings can extend in
die ^-direction. Therefore, the compressive stress along the .t-
direction is cancelled on both sides of the void. This result
causes highly compressive stresses along the jr-direction just
c
3
JO
1.
to
c
0)
520 525 530 535
Raman shift (enr 1 )
540
Fig. 5. A,Cro>mode micro-Raman spectra for various spots in the
cleaved (U00)r*N surface of ABLEG-GaN.
above the void. As a result, the wings expand in the in-
direction, and the deformation of the wings occurs. There-
fore, it is found that the wing deformation in the vicinity of
the coalescence boundary is due to the stress relaxation near
the tapered void.
The stress distributions can be characterized by the
frequency shift in die micro-Raman spectroscopy measure-
ments. 1 t]7 * Figure 5 shows the Aj(TO)-mode micro-Raman
spectra for various spots in the cleaved (1 100) Ga N surface in
the 6-um -thick ABLEG-GaN. The Raman peak wave number
depends on the measurement spot. The E2 mode lines showed
almost the same frequency shift as the Ai(TO)-mode. The
highest wave number was observed at the coalescence
boundary above the void (spot B), while the lowest was
observed at the bottom of the wing near the void (spot A). In
the seed region (spot C) and the midwing region (spot D), the
peak wave number was observed between spots A and B. The
frequency shift in the micro-Raman spectroscopy measure-
ments shows ihe existence of Ihe stress distribution in the
ABLEG-GaN films. The estimated stress difference between
spots A and B is about 200 MPa.' 61 These results are in good
agreement with the FEM calculation results.
PACE 9/16 * RCVD AT 12/23/2008 10:52:25 AM [Eastern Standard Time] • SVR:UBPTO-EFXRF-5/8 - DNIS: 2738300 ■ CSID:61 2-455-3801 * DURATION (mm-ss):05-40
12/23/2008 09:52
612.455.3801
HSML(JLS)
Page 10/16
Jpn. I. Appl. Phys. Vol. 42 (2003) Pi. 2. No. I OB
Tn summary, rwo- dimensional deformations and stress
distributions in ABLEG-GaN films have been studied by
AFM. two-dimensional FEM analysis, and micro-Raman
spectroscopy. The ABLEG-GaN wings slightly till, and the
direction of the wing tilt changes when the wings coalesce
with each other. After coalescence of the wings, the tilt angle
decreases with increase of the film thickness. By FEM
analysis and Raman spectroscopy, it has been revealed that
the deformation of the wings originates from the distributions
of thermal stress due to large mismatch of the thermal
expansion in the GaN seed layer and in the substrate. The
wing deformation decreases with increasing film thickness,
since the stress distribution becomes more uniform.
The author wishes to thank Dr. K. Eda for his encourage-
ment, and Drs. I. Kidoguchi and Y. Hasegawa for useful
discussion. The author wishes to acknowledge technical
support from H. Morita and E. Mizoguchi.
1) T. Nishinaga, T. Nakano and S. Zhand: Jpn. J. Appl. Phys. 27 (1988)
L964.
2) A. Usui, H. Sunakawa, A. Sakai and A. A. Yamaguchi: Jpn. J. Appl.
Phys. 36(1997) L899.
3) T. S. Zhetcva, O. H. Nam. M. D. Bremscr and R. F. Davis: Appl. Phys.
Lett. 7 1 (1997)2472.
A. ISHTBASPI ft al. L125I
4) S. Nafcamura, M. Senoh. S. Nagahama, N. Iwasa, T. Malsmhiu and T.
Mukai: MRS Internet J. Nitride Semicond. Res. 4SI (1999) Gl. I.
5) Y. Honda, Y. lyechika, T. Maed.i. H. Miyakc and K. Hiramaisu: Jpn. I.
Appl. Phys. 40 <2001>L309.
6) S. Tomiya. K. Punato, T. Asawuma, T. Hinn, S. Kijima, S. Asano and
M. Ikeda: Appl. Phys. Lett. 77 <2000> 636.
7) A. Sakai, H. Sunakuwa and A. U*ui: Appl. Phys. Lew. 73 (1998) 481.
8) t. Kidoguchi, A. Ishibashi. G. Sugaliara and Y. Ban: Appl. Phys. Lett.
7<>(2O0O) 3768.
9) A. Tshihashi, 1. Kidoguchi. C. Sngatiara and Y. Ban: J. Crysi. Growth
221 (2000) 338.
10) Q. Liu. A. Hoffmann, A. Kaschncr, C. Thomsca J. Christen, P. Veil
and R. Clos: Jpn. J. Appl. Phys. 39 (2000) L938.
1 1) S. Eint'eldt, A. M. RoskwwsVi, E. A. Preble and R. F. Davis: Appl. Phys.
Leu. 8ft (2002) 953.
12) SAMCEF 8.1 User 's Manuals (Samtech. Belgium and Surigikcn Co.
Ltd., Tokyo, 2000)
13) T. Detehprohm. K. Hiramatsu. K. Itoh and 1. Akasaki: Jpn. J. Appl.
Phys. 31 (1992)LJ454.
14) V. A. Savastenko and A. U. Sheleg: Phys. Status Solidi A 48 (1978)
KI35.
15) H. P. Maruska and J. J. Tietjen: Appl. Phys. Lett. 15 (1969) 327.
16) T. Kozawa, T. Kachi, H. Kano, H. Nagase, N. Koide and K. Manabe: J.
Appl. Phys. 77 (1995) 4389.
17) H. Siegle. P. Thnrian, L. Eckey. A. Hoffmann. C. Thom*en, B. K.
Meyer, H. Amano, I. Akasaki, T. Detehprohm and K_ Hiramatsu: Appl.
Phys. Lett. 68 0996) 1265.
PACE 10/16- RCVD AT 12/23/2008 10:52:25 AM (Eastern Standard TimeJ* SVR:USPTO-EFXRF-5/8 * DNIS:2738300 • CSID:612-455-3801 * DURATION (mm-ss): 05-40
12/23/2008 09:52
612.455.3801
HSML(JLS)
Page 11/16
CRYSTAL
GROWTH
ELSEVIER Journal of Crystal Growth 192 (1998) 73-78
Lateral overgrowth of high quality GaN layers on GaN/Al 2 0 3
patterned substrates by halide vapour-phase epitaxy
G. NataP, B. Beaumont, A. Bouille, S. Haffouz, M. Vaille, P. Gibart
Centre de Recherche sur VHeteroepitaxie etses Applications ( CRM EA -CN RS), Sophia Antipolis. G6560-Vatbonne> France
Received 20 February 1998
Abstract
The growth of GaN thick layers by halide vapour-phase epitaxy (HVPE) on metalorganic vapour-phase epitaxy
(MOVPE)-GaN/Al 2 03 substrates is reported. In a first step, we have shown that lateral overgrowth was enhanced
following preferential crystallographic directions. Double crystal X-ray diffraction (DCXRD) assessment in a) scan
showed full-width at half-maximum (F WHM) as small as 50 arcsec. On the way towards the realisation of self-supported
GaN substrates, the present study was extended to epitaxial lateral overgrowth (ELOG) on large surface GaN/Al 2 0 3
patterned substrates to achieve coalescence. Structural, electrical and optical characterisation of such layers was
performed, underlining the promising quality of these materials. © 1998 Elsevier Science B.V. All rights reserved.
PACS: 8i.05.Ea;81.15.Gh
Keywords: Selective epitaxy; Lateral overgrowth; Growth anisotropy; MOVPE; HVPE
1. IntroductioD
GaN and II1-V related compounds are essential
materials Tor devices operating in the green to UV
region of the spectrum. Blue LD's and green to blue
LED's have already been realised by the MOVPE
technique [1-7]. Presently, due to great Difficulties
in growing stable bulk GaN single crystals, the
major challenge is to overcome the lack of an
* Corresponding author. Fax: + 33 4 9395 8361; e-mail:
gn@crhea.cnrs.fr.
appropriate substrate. For example, Unipress in
Poland has grown GaN from gallium melts [8,9].
However, the size of GaN single crystals produced
by this technique is only a few square millimetres,
although their structural quality is very good.
Among the alternative substrates, the most wide-
ly used is A1 2 0 3 , although SiC, ZnO, MgAl 2 0 4 ...
have also been tried. However, all of these crystals
present large lattice parameters and thermal coef-
ficient differences with those of GaN leading to
highly strained and dislocated layers. Therefore, the
important defect densities always observed in GaN
layers, degrade tremendously the quality of GaN
based devices. During the last year, several groups
0022-0248/98/$ 19.00 & J998 Elsevier Science B.V. All rights reserved.
P1I: 50022-0248(98)004 1 3-B
PACE 11/16 * RCVD AT 12/23/2008 10:52:25 AM [Eastern Standard Time] * SVR:USPTO-EFXRF-5fc * DNIS:2738300 * CSID: 61 2-455-3801 ■ DURATION <mm-ss):05-40
12/23/2008 09:52
612.455.3801
HSML (JLS)
Page 12/16
74 G Natafet al. / Journal of Crystal Growth 1 92 (1998) 73- 78
have demonstrated the effectiveness of localised
epitaxial lateral overgrowth (ELOG) on patterned
substrates to reduce this defect density [10,11],
Most of the early works devoted to the elaboration
of GaN layers on sapphire substrates used the
HVPE process. More recently, HVPE which allows
high growth rates for good quality materials was
reported to supply thick GaN films on Al 2 0 3 sub-
strates [12,13]. In this work, HVPE has been car-
ried out directly, to produce ELOG layers up to
coalescence.
2. Experimental procedure
The growth of GaN was carried out in a home-
made HVPE system with a horizontal reactor.
A HQ -f N 2 mixture (purity 99.999%) was made to
react in a silica boat with gallium (purity
99.99999%) at 1000°C, then GaCl was introduced
with NH 3 in the growth zone to form GaN. N 2 was
used as the carrier gas. HC1 3 NH 3 and N 2 were
purified by means of getters. The heating phase was
carried out under NH 3 flow and the growth was
started as soon as the substrate temperature reach-
ed 1050°C. GaN layer thickness ranging from 30 to
150 urn were obtained at growth rates up to
30 um/h. Fig. 1 shows a schematic drawing of the
system.
The substrates used here were MOVPE-GaN
layers grown on (00 0 l)sapphire with a thin GaN
low-temperature buffer layer. Growth conditions
for these layers are published elsewhere [14]. GaN
layers were coated with a very I bin silicon nitride
film (~2 rim) in the same MOVPE run and then
the mask openings were defined by photolithogra-
phy and ECR etching.
In a first approach, two kinds of masks were
designed to assess the main' growth parameters.
The first one was star-shaped with line openings
Sum wide and 350 um long oriented with a 5°
angular increment. In one corner of this feature,
a rectangular field of hexagonal holes (5 um dia-
meter separated by 10 um) was also imple-
mented. The second one was made of a line array
with 5 or 10 um openings and a periodicity ranging
between 5 and 100 um. The stripe's length was
2 mm.
3. Growth results
Several growths with 0.5-3 h duration were per-
formed on each of these patterned GaN/Al 2 0 3
substrates. Fig. 2a shows a HVPE-GaN layer
grown on star-patterned openings. As a result of
growth on the star- patterned openings, an obvious
trend towards lateral extension for stripes following
the directions <0 1 1 0> can be observed, showing
that the lateral on vertical growth rate ratio is then
maximum (Fig. 2a). As pointed out by other
authors [15], the ends of the lines tend towards
a pyramidal form revealing the slow growing facets
{IT 01} (Fig. 2b). It must be pointed out that
the amount of lateral overgrowth is closely depen-
dent upon the concentration of atomic nitrogen
in the vapour phase, consequently it is magnified
with high temperature and high ammonia flows as
Ni+HCl
CiUiu
: Substrate
Exhaust
i
i
Mixing Zone
Fig. I. Schematic diagram of the HVPE reactor. The counter-flow NHj line allows to separate the mixing zone from that of GaCJ
introduction.
PACE 12/16 " RCVD AT 12/23/2008 10:52:25 AM (Eastern Standard TlmeJ * SVR:USPTO-EFXRF-5/8* DNIS: 2738300 " CSIO:61 2-455-3801 * DURATION (mm-ss):05^0
12/23/2008 09:52
612.455.3B01
HSML(JLS)
Page 13/16
G. Natafei al j Journal of Crystal Growth 192 (1998) 73- 78 75
<1 0 To>
«0 I10>
<- \ | 00> < 1 0 1 0>
Fig, 2. SEM microgaphs of a GaN layer grown on a star-pat-
terned substrate, (a) When the stripes follow the directions
<0 T 1 0>. LOG is enhanced; (b) detail of triangular termination
of growth following a stripe along <0 1 1 0>.
reported by Kapolnek ct ai. for MOVPE-grown
GaN [10].
For runs performed on stripe-patterned substra-
tes, coalescence was achieved for stripes following
<0 T 1 0> directions, when window openings were
5 or 10 um wide, and, provided the spacing region
was not more than 10 jim. Fig. 3a represents a
SEM cross-section micrograph of a continuous
GaN film grown on a stripe-patterned region. In
the region near the interface, the voids indicate the
location where the coalescence took place. Emerg-
ing from each void, a chimney goes vertically up to
the surface. For dielectric width larger than 10 um,
hexagonal holes appeared even for growth times as
long as 3 h showing the difficulty for GaN bridges
to completely link perpendicular to the stripes di-
rection (Fig. 3b). The mechanism involved in such
lateral overgrowth appears quite different from the
one observed for other IJI-V semiconductors as
GaAs [16]. Here there is not an expansion of a lat-
eral plane leading to coalescence, but rather local-
ised nucleations throwing separate lateral bridges
between two stripes, followed by an extension
along the stripes to form a continuous film
(Fig. 3c).
Concerning HVPE-GaN growths on hexagonal
holes, two observations arose. First, at the early
30min of the growth, well-shaped hexagonal
prisms were formed (Fig. 4a) whose facets belong to
the { 1 T 0 I } family as already seen by the MOVPE
technique [17]. For a longer growth time, bridges
appeared linking the different pyramids and then
coalescence could take place (Fig. 4b). The result
was a flat homogeneous film with remaining holes.
Preliminary assessment by DCXRD of such small
areas of GaN overgrown layers showed a FWHM
of 50 arcsec in a scan.
4. HVPE lateral overgrown GaN
The next step towards the achievement of self-
supported substrates was the realisation of GaN
lateral overgrown thick layers on larger surfaces.
For this purpose, two other masks were designed to
allow large area patterns, consisting either of hex-
agonal dots or of parallel stripes. At present, only
the first one was tried. Using previous growth con-
ditions, it was possible to realise continuous
laterally overgrown thick GaN layers (up to
150 um) presenting a flat surface. However, al-
though globally very smooth, the surface presented
numerous hexagonal holes with a depth up to
50 um, depending on the way the linking between
neighbouring hexagonal pyramids took place.
5. Structural assessment
DCXRD measurements carried out on lateral
overgrown HVPE-GaN layers on large surface
hexagonal patterned substrates showed FWHM in
co scan ranging between 300 and 450 arcsec. These
results are clearly poor as compared to the 50 ar-
csec previously reported in this paper for ELOG on
a very small area. Recently, Sakai et al. [17]
showed, by TEM observations on selectively grown
PACE 13/16 * RCVD AT 12/23/2008 10:52:25 AM [Eastern Standard Time] ■ SVR:USPTO-EFXRF-5/8 * DNI8:2738300 * C8ID:61 2-455*3801 * DURATION (mm-SS): 05-40
12/23/2008 09:52
612.455.3B01
HSML(JLS)
Page 14/16
76
G Natqfet at. /Journal of Crystal Growth 192 (1998) 73-78
Fig, 3. SEM micrographs of GaN grown on a stripe-patterned substrate, (a) The cross section shows voids terminated by a chimney at
the interface between the MOVPE layer and the LOG-HVPE layer; <b) surface view evidences that complete coalescence is achieved on
stripes with lower periodicity (Sum maskx5ym window at the right bottom); (c) on stripes with higher periodicity (100 nm
mask* 10 um window) bridges showing the LOG mechanism are seen.
GaN by HVPE on MOVPE layers, that most of
the threading dislocations vertically aligned in the
MOVPE-grown layer, propagated laterally around
the Si0 2 mask in the HVPE film, before the film
thickness reached about 5 um. Hence, this enlarge-
ment of FWHM in the case of large area ELOG
GaN, could be attributed to the sample bending
due to the important residual stress existing be-
tween the GaN layer and the Al 2 0* substrate
rather than a degradation of the defect density.
Consequently, we need to check the defect density
observed on such a layer by TEM direct observa-
tion. This study will be carried out in the near
future.
6, Physical characterisation
Hall- Van der Pauw assessment performed on
ELOG-GaN showed room temperature mobility
up to 210cm 2 /Vs for electron concentrations as
low as 3 x 10 11 cm" 3 .
Low-temperature (10 K) photoluminescence
spectrum (Fig. 5), is dominated by near band edge
excitonic transitions involving free exciton
A (3.481 eV), a donor bound exciton l 2 (3.474 eV)
and acceptor bound exciton /, (3.462 eV). In the
3.27 eV range, donor-acceptor pairs recombina-
tions (D°A°) were observed involving the usual
donors and acceptors of GaN. According to the
literature, these numerical values show that this
GaN layer is fully relaxed. Closer examination of
this spectrum showed that the intensity ratio of
band edge to acceptor related transitions got close
to two orders of magnitude. This is in agreement
with the low residual impurity level measured by
the Hall effect. According to previous studies per-
formed on HVPE-GaN layers on sapphire, there
was no yellow band detected in these ELOG-GaN
samples.
PAGE 14/16 » RCVD AT 12/23/2008 10:52:25 AM [Eastern Standard Tlmel* SVR:USPTO-EFXRF-5/8 " DNIS:2738300 • CSID:91 2-455-3801 ■ OURATION (mm-ss): 05-40
12/23/2008 09:52
612.455.3B01
HSML (JLS)
Page 15/16
G. Natafet a!. /Journal of Crystal Growth 192 (1998) 73-78
Fig. 4. SEM micrographs of GaN grown on an hexagonal holes
field, (a) After 30 tnn growth, well shaped pyramids are formed;
(b) later, coalescence takes place leading to flat surfaces separ-
ated by holes or valleys.
«2
3fl V V M XB &
Fig. 5. Low-temperature photoluminesccnce spectrum. Near
band edge feature shows I 2 (3.4746 eV), A (3.4809 eV) and
l { (3.4620 eV> lines.
7. Discussion
Thick GaN layers overgrown by HVPE on pat-
terned MOVPE-GaN/Al 2 0 3 substrates present
structural and physical properties of high quality.
The total reduction in the density of defects has
presently to be demonstrated and further progress
is still expected but it is clear that large area ELOG
represents an important step towards self-sup-
ported GaN substrates.
Acknowledgements
This work was supported by an ESPRIT-LTR
contract from the European Union, LAQUANI
N°20968. The authors wish to thank MM J.C
Guillaume, M. Passerel, Dr. M. Leroux and Dr. O.
Parillaud for helpful advice.
References
[1] S. Nakamura, C. Fasol, The Blue Laser, Springer, Berlin,
1997.
[2] R.P. Vaudo, ID. Gocpfert. T.D. Moustakas, D.M.
Beyea, TJ. Frey, K. Meehan, J. Appl. Phys. 79 (5) (1996)
2779.
[3] T.J. Schmidt, X.H. Yang, W. Shan, JJ. Song, W. Kim. 0.
Aktas, A. Botchkarev, H. Morkoc, Appl. Phys. Lett. 68(13)
(1996) 1820.
[4] S. Kurai, Y. Naoi, T. Abe, S. Ohmi, S. Sakai, Jpn. J. Appl.
Phys. 35(1996) L77.
[5] P.A. Maki, R.J. Molnar, R.L. Aggarwal, Z.L. Liau,
I, Melngailis. Mater. Res. Soc Symp. Proc 395 (1996)
919.
[6] B. Beaumont, P. Gibart, M. Leroux, E CaDeja, E Munoz,
ErMRS 97, Strasbourg, 1997, Mater. Set Eng., to be pub-
lished.
[7] a Beaumont, F. Calle, S. Haffouz, E. Monroy, M. Leroux,
E Calleja, P. Lorenzini, E. Munoz, P. Gibart, Proc. Int.
Conf. On Silicon Carbide, Ill-Nitrides and Related Mater-
ials. Stockholm, to be published.
[8] S. Porowski, J. Crystal Growth 166 (1996) 583.
[9] T. Suski, P. Perlin, M. Leszczynski, H. Tcisseyre, T.
Grzegory, J. Jun, M. Bockowski, S. Porowski. K. Pakula,
A. Wysmolck, J.M. Baranowski. Mater. Res. Soc. Symp.
Proc. 395(1996) 15.
[10] D. Kapolnek, S.' Keller, R. Vetury, R.D. Underwood. P.
Kozodoy, S.P. Den Baars, U.K. Miahra, Appl. Phys. Lett.
71 (9) (1997) 1204.
[I I] X. Li, A.M. Jones, S.D. Roh, D.A. TurnbuU, S.G. Bishop,
J J. Coleman, J. Electron. Mater. 26(3) (1996) 306.
PACE 15/16 * RCVD AT 12/23/2008 10:52:25 AM [Eastern Standard Tlmel » SVR:USPTO-EFXRF T 5/8 • DNIS:2738300 * CSID:61 2-455-3801 * DURATION (mnvss):05-40
12/23/2008 09:52
612.455.3801
HSML(JLS)
Page 16/16
78 G. Nataf et at. / Journal of Crystal Growth 1 92 (1 998) 73- 78
[12] K. Naniwae, S. Itoh, H. Amano, K. Itoh, K. Hiramatsu, L
Akasaki, J. Crystal Growth 99 (1990) 381.
[13] T. Detchprohm, K. Hiramatsu, H. Amano, I. Akasaki,
Appl. Phys. Lett. 61 (22) (1992) 2688.
[14] B. Beaumont, M. Vailie, T. Boufaden, B. El Jani, P. Gibart,
J. Oystal Growth 170 (1997) 316.
[15] Ok'Hyun Nam, M.D. Bremser, B.L. Ward, R.J. Nemaruch,
R.F. Davis, Jpn. J. AppL Phys. 36(1997) L 532.
[16] G. NataC M. Leroux, S.M. Laugt, P. Gibart, J. Crystal
Growth 165 (1990) I.
[17] A Sakai, H. Sunakawa, A. Usui, Appl. Phys. Lett. 71 (16)
(1997) 2259.
PACE 16/18 • RCVD AT 12/23/2008 10:52:25 AM [Eastern Standard TimeJ* SVR:USPTO-EFXRF-5/8* DNIS:2738300 " CSID:61 2-455-3801 * DURATION <mm-ss):05-*0